Comparative analysis of multi-resolution optical flow techniques for flow rate estimation

In April 20, 2010, the largest amount of oil in history leaked into the Gulf of Mexico (GoM) due to the oil rig explosion. For the fact of the lack of accurate technique for estimating the volume of oil spilled in this incident, ad-hoc techniques were employed from other fields e.g. Particle Image Velocimetry (PIV) from fluid dynamic experiment. Of the different techniques employed, the technique by Crone based on optical plume velocimetry was the most accurate. However, the accuracy of estimation differs depending on the choice of algorithms. In this work, a comparative analysis of optical flow algorithms based on image warping theory and on Lucas and Kanade´s algorithm was performed. Crone´s experimental setup was replicated and a video camera was used to record the flow. The two algorithms were used to estimate flow rate. To improve estimation accuracy, a multi-resolution framework was used. Optical flow based on image warping outperforms that based on Lucas and Kanade in representing the velocity fields in all regions of flow. Moreover, a comparison between the estimated flow rates with actual flow rates resulted in a non-zero crossing of 0.064 and 0.128 m/s for image warping and Lucas and Kanade respectively.